The DeepMind Fast Language Learning Tasks is a set of machine-learning tasks that requires agents to learn the meaning of instruction words either slowly (i.e. across many episodes), quickly (i.e. within a single episode) or both.

The tasks in this repo are Unity-based.

These tasks are provided through pre-packaged Docker containers.

This package consists of support code to run these Docker containers. You interact with the task environment via a dm_env Python interface.

Please see the documentation for more detailed information on the available tasks, actions and observations.

dm_fast_mapping requires Docker, Python 3.6.1 or later and a x86-64 CPU with SSE4.2 support. We do not attempt to maintain a working version for Python 2.

Note: We recommend using Python virtual environment to mitigate conflicts with your system's Python environment.

Download and install Docker:

• For Linux, install Docker-CE
• Install Docker Desktop for OSX or Windows.

You can install dm_fast_mapping by cloning a local copy of our GitHub repository:

$ git clone https://github.com/deepmind/dm_fast_mapping.git
$ pip install ./dm_fast_mapping

You can install the dependencies for the examples/ with:

$ pip install ./dm-fast-mapping[examples]

Once dm_fast_mapping is installed, to instantiate a dm_env instance run the following:

import dm_fast_mapping

settings = dm_fast_mapping.EnvironmentSettings(seed=123,
    level_name='fast_slow/fast_map_three_objs')
env = dm_fast_mapping.load_from_docker(settings)

If you use dm_fast_mapping in your work, please cite the accompanying paper:

@misc{hill2020grounded,
      title={Grounded Language Learning Fast and Slow},
      author={Felix Hill and
              Olivier Tieleman and
              Tamara von Glehn and
              Nathaniel Wong and
              Hamza Merzic and
              Stephen Clark},
      year={2020},
      eprint={2009.01719},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

For the with_distractors tasks, please also cite the source for those tasks:

@misc{lampinen2021towards,
      title={Towards mental time travel:
             a hierarchical memory for reinforcement learning agents},
      author={Lampinen, Andrew Kyle and
              Chan, Stephanie C Y and
              Banino, Andrea and
              Hill, Felix},
      archivePrefix={arXiv},
      eprint={2105.14039},
      year={2021},
      primaryClass={cs.LG}
}

This is not an officially supported Google product.